The task
The two articles below take opposite positions on one question:
Should Harrow Casualty require its claims analysts to work on site five days a week?
Read both. Then analyze both positions to determine which one is better supported by evidence. Use specific evidence from both articles in your answer.
You are not being asked which side you agree with, and you will not be scored on which one you choose. You are being asked to judge the arguing — and either article can be argued for. Allow about 45 minutes.
Article 1
The Room Teaches
1I want to argue for the room, and I want to do it with our own records rather than with a feeling about how things used to be.
2A new claims analyst is considered trained when she can carry a full caseload of 60 open files without a senior analyst reviewing her work. Analysts hired between 2015 and 2019, who trained on site, reached that point in a median of nine weeks. Analysts hired between 2021 and 2024, who trained remotely, reached it in a median of sixteen.
3Last fall the department surveyed all 38 of its supervisors; every one of them answered. Thirty said remote new hires ask fewer questions than on-site new hires did, and that the questions they do ask tend to arrive after the mistake rather than before it. Eleven wrote in some version of the same comment without being prompted: that they now find out a new analyst has been doing something wrong when the audit finds it.
4There is a reason for that, and it is not discipline. A new analyst sitting in a row of desks hears the person beside her talk a furious claimant down from a ledge. She hears the phrasing. She hears where the voice slows. Nothing in our training manual contains that, and nobody schedules a video call to transmit it. It is overheard, or it is not learned.
5Nobody here is asking anyone to commute for its own sake. We are asking for the room in which the work is learned, because we have watched what happens to people who have to learn it alone.
Article 2
Count the Work, Not the Chairs
1Mr. Brennan and I are describing the same department. We are counting different things.
2Claims closed per analyst per month rose from 41 in 2019 to 45 in 2024, a gain of about nine percent. Over the same five years the department's error rate, audited every quarter by an outside firm the company does not choose and cannot fire, held at about two percent. It did not drift up in 2021 and it has not drifted up since. The work got done, it got done on time, and it got done correctly, by people at kitchen tables in Jackson Heights and Canarsie.
3Turnover tells the same story from the other end. Of the analysts on the remote-eligible track, four percent left the company last year. Of the analysts required on site — the intake and mailroom units — nineteen percent left.
4There is also a cost that never appears on a company document. Staff who live outside Manhattan average 74 minutes each way and $145 a month in fares. Five days a week, that is roughly twelve unpaid hours a week per person, borne entirely by the employee.
5And I will grant Mr. Brennan his sixteen weeks. If new analysts take longer to come up to speed, then fix the training: pair every new hire with a senior for a named period, put the first six weeks on site, and measure whether it closes the gap. That is a targeted answer to a measured problem. Moving two hundred people's lives to solve a training problem is not.
Your 45 minutes
Stop here if you have not written yet. What follows names the evidence in both articles and shows where each one is weak. Read it before you write and you will not be practicing the thing the test measures — which is finding all of this yourself, in forty-five minutes, with nobody helping.
After the clock — what was actually on the table
What each article put in front of you
The Room Teaches
- An outcome measure with a stated threshold. “A full caseload of 60 open files without review” — he defines what “trained” means before he measures it. That is more than most arguments do.
- A median, not an average. Nine weeks against sixteen. Using the median means one disastrous trainee cannot drag the number.
- A survey of a whole population. All 38 supervisors, 30 agreeing — not a sample.
- A mechanism, told as a scene. Paragraph 4 explains why the gap might exist.
Count the Work, Not the Chairs
- Output and quality together. 41 to 45 claims a month, with the error rate flat. Either number alone could be explained away; the pair is harder to dismiss, because producing more and making no more mistakes is the thing an employer says cannot happen.
- An independent auditor. “An outside firm we do not choose” — she names who checked and whose it is not.
- A retention comparison with two groups. Four percent against nineteen.
- A cost the company's records do not carry. 74 minutes, $145 a month, twelve unpaid hours a week.
- A concession and a targeted remedy. She grants his central finding and proposes a test for it.
Where each one is thin
Neither article is the better one. Each has evidence a careful reader can push on, and naming that pressure — in specific words, about a specific sentence — is most of what the first scoring trait rewards.
The Room Teaches
- The two groups are in different years, and the years are not alike. 2021 to 2024 remote training was improvised in an emergency; 2015 to 2019 on-site training was a mature routine. His comparison does not separate remote from invented in a hurry, which is the whole question.
- The supervisors are being asked about their own relevance. Thirty of 38 people whose job is supervising report that people are harder to supervise from a distance. That is real information, and it is not disinterested.
- Paragraph 4 is an assertion with a story attached. A vivid scene is not a measurement. Notice that it is the only paragraph in his passage with no number in it, and that it carries the passage's central claim.
- He never says whether the department's output changed. He measures how fast new people got trained and says nothing about what the department produced — which is the number Ranganathan reaches for.
Count the Work, Not the Chairs
- Claims closed per analyst can move for reasons that have nothing to do with where anyone sat. If the simplest claims were automated between 2019 and 2024, the remaining files would take longer and the count would fall, not rise — or the reverse, if routing changed. She does not say whether the mix of claims stayed the same.
- The retention comparison compares two different kinds of job. The on-site group is intake and mailroom. Those roles have higher turnover than analyst roles in almost every company, remote or not. This is the weakest number in her passage and it is presented as one of her strongest.
- The commute cost is a cost, not evidence about the work. It belongs in the argument. It does not answer Brennan, who is talking about how quickly people learn.
- She grants the sixteen weeks and does not test it. Her remedy in paragraph 5 is a proposal, offered with no evidence that pairing or six weeks on site would close the gap.
Two essays, either of which could score well
If both of those look like real essays to you, the set is working as intended. On the test you will have two articles built the same way, and no time to find the perfect answer, because there isn't one. Pick the side you can support in forty-five minutes and support it.
Check your own
The Extended Response is scored on three traits. Read your draft against them, one at a time — and then read How the essay is scored, which has the full rubric and two responses to the same task, one earning a 6 and one earning about a 2.
- Analysis of argument and evidence. Did you say which article is better supported, in a sentence, near the beginning? Did you quote or name specific evidence from both articles — not just the one you chose? Did you say what is wrong with the evidence on at least one side, rather than only what is right?
- Development and organization. Does each paragraph do one job? Does a reader who has not seen the articles know what they said? Did you get to the end, or did you run out of clock in the middle of a paragraph?
- Clarity and command of English. Read it aloud. Sentences that you stumble over reading are sentences a scorer stumbles over too. Check that every sentence has a subject and a verb and ends where it should.
One more, which is not on the rubric and costs you the most: did you argue about which side is right instead of which side is better supported? That essay can be beautifully written and still score low, because it answers a question nobody asked.